Previous studies that have examined the effects of specific aspects of 12-step participation and social network composition on abstinence have focused mostly on alcohol-related outcomes and have screened out drug dependent persons. This article explores whether these predictors differentially affect abstinence based on DSM-III-R substance dependence disorder (alcohol dependence, drug dependence, and both alcohol and drug dependence). A heterogeneous community sample of treatment seekers (N=302) randomized to day treatment programs were followed at 6 and 12 months. Bivariate and multivariate regression models were used to test whether engagement in 12-step practices and social network influences to drink or use drugs predicted total abstinence from alcohol and drugs differentially by dependence disorder. Chi-square automatic interaction detector (CHAID) segmentation analyses were then conducted to identify the specific 12-step activities and social network thresholds that best distinguished higher rates of abstinence in each dependence category. Results showed that the number of 12-step meetings attended and number of prescribed 12-step activities engaged in similarly predicted abstinence for alcoholics, drug addicts, and those dependent on both alcohol and drugs. However, specific activities were associated with abstinence differentially by dependence disorder. While many activities differentiated abstinence for drug addicts and those dependent on both alcohol and drugs, for alcoholics only two Alcoholics Anonymous (AA) activities distinguished abstinence (having a sponsor and doing service). Key predictors of abstinence (CHAID) varied by follow-up and dependence disorder, except for doing service in AA and/or Narcotics Anonymous, which was the only specific 12-step activity that was a best predictor of abstinence in all three categories one year following treatment. Thus, "giving back" to one's peer community through service work, an important 12-step belief, seems to be universally valuable later in recovery. As for social network influences, a multivariate regression model showed that having a higher proportion of abstinent individuals in the network was associated with abstinence for alcoholics at 6 months only and for drug dependent persons at 12 months only. CHAID models supported these results and provided specific thresholds for 12-step measures (e.g., >20 meetings for alcoholics, 2 or more nondrinkers in the social network, 3 or more persons supporting reduction for those dependent on both alcohol and drugs, and having 2 or more nondrinkers for those dependent on drugs only). These results support the value of treatment providers prioritizing certain 12-step-related practices and social network changes based on their client dependence profiles. Early on, those with an alcohol diagnosis need to make a commitment to meetings and obtain a sponsor; also, they need to place themselves in a network that encourages sobriety. Early on, those who are drug-dependent-only especially need to become connected with 12-step programs to the extent that they consider themselves a member, and, later, saturate themselves in a highly supportive and predominantly nondrinking environment. Alcohol and drug dependent clients need more intense ongoing 12-step involvement (sponsor and meetings) as well as having nondrinking individuals and people supportive of abstinence in their network. For all clients, doing service is especially important at the longer 12-month posttreatment timeframe.